correlates of bushmeat in markets and depletion of wildlife

Contributed Paper

Correlates of bushmeat in markets and depletionof wildlife

John E. Fa,∗†‡ Jesus Olivero,† Miguel A. Farfan,†§ Ana L. Marquez,† Jesus Duarte,†§Janet Nackoney,∗∗ Amy Hall,∗ Jef Dupain,†† Sarah Seymour,∗ Paul J. Johnson,‡‡David W. Macdonald,‡‡ Raimundo Real,† and Juan M. Vargas†∗Durrell Wildlife Conservation Trust, Les Augres Manor, Trinity, Jersey, JE3 5BP, United Kingdom†Universidad de Malaga, Grupo de Biogeografıa, Diversidad y Conservacion, Departamento de Biologıa Animal, Facultad de Ciencias,Campus de Teatinos s/n, 29071 Malaga, Spain‡ICCS, Division of Biology, Imperial College London, Ascot SL5 7PY, United Kingdom, email [email protected]§Biogea Consultores, C/Navarro Ledesma no 243, Malaga, 29010, Spain∗∗Department of Geographical Sciences, University of Maryland, College Park, MD, 20740, U.S.A.††African Wildlife Foundation Conservation Centre, 00502, Nairobi, Kenya‡‡Wildlife Conservation Research Unit (WildCRU), Zoology Department, University of Oxford, Abingdon, OX13 5QL,United Kingdom

Abstract: We used data on number of carcasses of wildlife species sold in 79 bushmeat markets in a region ofNigeria and Cameroon to assess whether species composition of a market could be explained by anthropogenicpressures and environmental variables around each market. More than 45 mammal species from 9 orderswere traded across all markets; mostly ungulates and rodents. For each market, we determined median bodymass, species diversity (game diversity), and taxa that were principal contributors to the total number ofcarcasses for sale (game dominance). Human population density in surrounding areas was significantly andnegatively related to the percentage ungulates and primates sold in markets and significantly and positivelyrelated to the proportion of rodents. The proportion of carnivores sold was higher in markets with highhuman population densities. Proportion of small-bodied mammals (<1 kg) sold in markets increased ashuman population density increased, but proportion of large-bodied mammals (>10 kg) decreased as humanpopulation density increased. We calculated an index of game depletion (GDI) for each market from the sumof the total number of carcasses traded per annum and species, weighted by the intrinsic rate of naturalincrease (rmax) of each species, divided by individuals traded in a market. The GDI of a market increased asthe proportion of fast-reproducing species (highest rmax) increased and as the representation of species withlowest rmax (slow-reproducing) decreased. The best explanatory factor for a market’s GDI was anthropogenicpressure—road density, human settlements with >3000 inhabitants, and nonforest vegetation. High and lowGDI were significantly differentiated by human density and human settlements with >3000 inhabitants. Ourresults provided empirical evidence that human activity is correlated with more depleted bushmeat faunasand can be used as a proxy to determine areas in need of conservation action.

Keywords: bushmeat harvest, favorability function, game depletion, indexes, mammals

Correlaciones entre los Mercados de Carne de Animales Silvestres y la Disminucion de la Vida Silvestre

Resumen: Usamos datos numericos de carcasas de especies de vida silvestre vendidas en 79 mercadosde carne de animales silvestres, en una region de Nigeria y Camerun, para evaluar si la composicion deespecies de un mercado puede explicarse por medio de presiones antropogenicas y variables ambientalesalrededor de cada mercado. Mas de 45 especies de mamıferos de nueve ordenes fueron vendidas en todoslos mercados, la mayorıa ungulados y roedores. Para cada mercado determinamos el promedio de masacorporal, diversidad de especies (diversidad de presa) y taxones que principalmente contribuyeron al numerototal de carcasas en venta (dominancia de presa). La densidad de poblacion humana en las areas colindantesestuvo negativa y significativamente relacionada con el porcentaje de ungulados y primates vendidos en

Paper submitted October 3, 2013; revised manuscript accepted September 21, 2014.

1Conservation Biology, Volume 00, No. 0, 1–11C© 2015 Society for Conservation BiologyDOI: 10.1111/cobi.12441

2 Bushmeat and Depletion of Wildlife

los mercados y positiva y significativamente relacionada con la proporcion de roedores. La proporcion decarnıvoros vendidos fue mas alta en los mercados con altas densidades de poblacion humana. La proporcionde mamıferos de pequeno tamano (< 1kg) vendidos en los mercados se incremento conforme se incrementabala densidad de poblacion humana, pero la proporcion de mamıferos de gran tamano (> 10kg) disminuyoconforme aumento la densidad de poblacion humana. Calculamos un ındice de disminucion de presas (IDP)para cada mercado, a partir de la suma del numero total de carcasas vendidas por ano y especie, sopesadopor la tasa intrınseca de incremento natural (rmax) de cada especie, dividida por el numero de individuosvendidos en un mercado. El IDP de un mercado se incremento conforme lo hizo la proporcion de especies derapida reproduccion (rmax mas alta) y conforme disminuyo la representacion de especies con la rmax mas baja(de baja reproduccion). El mejor factor explicativo para el IDP de un mercado fue la presion antropogenica –densidad de carreteras, asentamientos humanos con mas de 3000 habitantes y vegetacion no boscosa. Los IDPaltos y bajos fueron significativamente diferenciados por la densidad humana y los asentamientos humanoscon mas de 3000 habitantes. Nuestros resultados proporcionaron evidencia empırica de que la actividadhumana esta correlacionada con la disminucion de especies cazadas y puede usarse como indicador paradeterminar areas que necesiten acciones de conservacion.

Palabras Clave: cosecha de carne de animales silvestres, disminucion de presa, funcion de favorabilidad, ındices,mamıferos

Introduction

Wildlife for sale in bushmeat markets is likely to reflectboth the abundance of the species in the wild and howhunters adjust their strategy as the ecosystem becomesdepleted of the more profitable species (Fa 2007). Al-though the type of habitat affects the composition ofprey communities (as intact forest is transformed intoagricultural or open land), the composition of species andnumber of carcasses appearing in markets in a specificlocality could indicate an overall defaunation resultingfrom the effects of increased hunting pressure (Dirzo &Miranda 1991).

Collection of reliable information on bushmeat offtakeover large regions, gathered within the same period, ishighly resource intensive; hence, few such studies havebeen undertaken (but see Fa et al. 2006). As a result,sufficient data to support effective management deci-sions at national and regional levels are lacking. However,because bushmeat markets are found in many villages,towns, and cities throughout West and Central Africa,data on animals traded within them can be used as a rapidassessment tool to measure faunal extraction at regionalscales (Fa 2007). Although there are caveats to the inter-pretation of data gathered at bushmeat markets, changesin species profiles and prey volume at different tradingpoints can permit a broad understanding of wild meatextraction across landscapes. This can allow managers toidentify potentially overhunted and less-disturbed catch-ment areas for consideration in strategic planning.

Most markets in West and Central Africa are dominatedby the sale of ungulates, rodents, and primates and,less so, by carnivores (Fa 2000). The composition ofspecies for sale in bushmeat markets is influenced bythe local history of hunting (Cowlishaw et al. 2005)because vulnerable taxa (slow reproducers such as largeungulates and primates) are depleted first and replaced

by smaller-bodied robust taxa (fast reproducers) suchas rodents and small antelopes. Consequently, increasedprominence of species with high reproductive potential(as defined by the intrinsic rate of natural increase rmax)sold in bushmeat markets characterizes heavily exploitedcatchments (Dupain et al. 2012).

We investigated the relationship between environ-mental characteristics and anthropogenic pressures incatchment areas and the species appearing in bushmeatmarkets. We tested the hypothesis that the percent com-position of various mammal groups (ungulates, rodents,primates, and carnivores) in a market is a crude measureof depleted faunas of the areas supplying that market.We utilized a database of traded wild animals in 79 ruralmarkets from southern Nigeria and Cameroon (Fa et al.2006; Macdonald et al. 2011, 2012). We also developedan index of game depletion to distinguish betweenmarkets in depleted and nondepleted catchments. Wedetermined anthropogenic pressures and environmentalconditions favoring faunal depletion.

Methods

Study Area

The study area (35,324 km2) was at the left bank of theCross River in southeastern Nigeria (10,795 km2) andextended as far south as the Sanaga River in Cameroon(24,529 km2) to around 200 km inland (Fig. 1). The mainvegetation type is humid tropical forest. At higher eleva-tions there is montane forest and other plant formations.Topographic relief is low at the eastern and westernmargins but increasingly rugged in the foothills of theNigerian–Cameroonian Mountains. Total annual rainfallin the region is between 2000 and 3500 mm, but thereis great intraregional variation related to proximity tothe coast and elevation and compounded by rain-shadow

Conservation BiologyVolume 00, No. 0, 2015

Fa et al. 3

Figure 1. Locations of studymarkets in theCross-Sanaga Rivers region,Nigeria, and Cameroon andland-cover and land-usedata from GlobCover LandCover version 2.2 databasefor 2005–2006 (Bicheron etal. 2008).

effects; the wet season lasts from April to October. Hu-midity is constantly high (>70%), and temperatures rangefrom 15 ºC to 33 ºC. This region contains the world’spopulations of the Cross River gorilla (Gorilla gorilladiehli), Nigeria-Cameroon chimpanzee (Pan troglodytesvellerosus), and mainland drill (Mandrillus leucophaeusleucophaeus) (Oates et al. 2004). The forest elephant(Loxodonta cyclotis) also occurs in the region.

Forest conversion remains high in the region and hashad considerable impacts on larger mammals (Forbosehet al. 2007). There are a number of protected areas in theregion, the largest of which is the Cross River NationalPark (3586 km2, divided in 2 parts Oban and Okwangwo)in Nigeria and the Korup National Park (1256 km2) in

Cameroon. An estimated 5.2 million people lived withinthe study area (1.3 million in Nigeria and 3.9 million inCameroon). The majority of inhabitants were concen-trated in populous conurbations (Calabar and Douala)and their municipalities. In Nigeria 33% of the study area’spopulation lived in Calabar, while in Cameroon 37% livedin Douala (Fa et al. 2006).

We surveyed bushmeat-trading points in the study re-gion from August 2002 to January 2003. We refer to thesepoints as markets. Although these markets may accountfor a great majority of the bushmeat trade taking placein the study area, they do not capture all of it becausehunters sometimes sell to consumers directly. We sam-pled 79 markets (37 in Nigeria and 42 in Cameroon).



Market Data Collection

Twelve local field assistants (5 in Cameroon, 7 in Nigeria)were trained and managed by 2 in-country managers (S.Seymour in Nigeria and J. Dupain in Cameroon). Eachfield assistant was responsible for overseeing a group ofbushmeat markets (4 to 9 each), where local reporters(an average of 3 reporters per market, total 250) collecteddata. Assistants recruited, paid, and monitored local re-porters. They also assembled and checked data sheets.

At each market, we recorded the identity of the vendorand the species and condition (dried or smoked or freshor alive) of whole or parts of animals for sale. Amphib-ians, reptiles, birds, and mammals were recorded in oursampled markets, but over 95% of all specimens weremammals (Fa et al. 2006). We used only data on mammalspecies. Although most animals sold were whole, bodyparts (mostly limbs) were sometimes found in market.In the latter case, we estimated the minimum numberof individuals (MNI) as the fewest possible number ofanimals in an assemblage of body parts. If there were 2legs, a left and a right, then the MNI = 1. If there were2 left legs then the MNI = 2. Only carcasses belongingto the family Sciuridae, the species Phataginus tricuspisand Uromanis tetradactyla, and some individuals of thefamily Herpestidae were unreliably identified. In thesecases, we grouped them together for analytical purposes.

A mean of 142.3 days (SE 5.0, range 29–148) weresampled per market in Nigeria and 152.2 days (SE 1.40,range 100–167) in Cameroon. This meant a total of 7594market days were sampled (4936 days in Nigeria and 2658days in Cameroon).

The annual number of animals of each mammal speciestraded per market was calculated as the mean number ofcarcasses (either dried or smoked or fresh or alive) soldper market day (sample days during which carcasses werecounted including those where none were registered)multiplied by 365 days. Within-year variation in numberof animals traded was not considered because we weredescribing annual trade per market.

Anthropogenic Pressures and Environmental MeasurementVariables

Empirical studies show that population densities arelower in hunted versus unhunted areas and that com-monly the density of large-bodied species increases asdistance from hunter settlements increases (Nasi et al.2011). We measured the prevailing environmental and an-thropogenic characteristics withina 15-km radius (here-after, buffer) surrounding each market. We based the sizeof the buffer on known distances traveled by hunters(n = 72) from their villages in Equatorial Guinea (medianof 15.25 km, range 8.5–30 km) in Kumpel (2006).

We used only markets in rural settlements. Foreach market buffer, we used 10 anthropogenic and

6 land cover and use type variables to describe thesite’s main environmental characteristics and humanpressures (Supporting Information). We determinedhuman population distribution (density and numberof settlements of >3000 inhabitants) as a proxy forhunter numbers and relative bushmeat demand; humanaccessibility (roads and navigable river networks);and agricultural activity and conversion (active fires).Environmental factors included land-use and land-covercategories (Supporting Information).

Bushmeat Market Game Profiles

For each market, we estimated the general characteristicsof the assemblage of species for sale (Supporting Informa-tion) with a game diversity and game dominance index(Vargas et al. 2004).

Because weights of individual carcasses could not beobtained for all sites, we extracted the average body massfor every traded species (or closest congeneric) from Faand Purvis (1997) to standardize comparisons betweensites (Supporting Information). From these, we calculatedthe median body mass of all species sold in each market.The annual biomass traded in a market was estimatedfrom the total number of carcasses of each taxa sold ina year multiplied by the mean body mass of the corre-sponding species.

We used the intrinsic rate of natural increase, rmax, ofeach species as a means of distinguishing between fastand slow breeding, a predictor of extinction proneness(Rowcliffe et al. 2003). Allometric regression was usedto predict rmax values based on body mass (Support-ing Information). Thus, if the number of individuals ofspecies with an elevated rmax in a market sample is high,than there is a preponderance of fast-breeding speciesremaining in the market catchment area and therefore adepleted faunas (i.e., faunas that have lost most of thelarge-bodied taxa [Cowlishaw et al. 2005; Dupain et al.2012]). We derived game depletion index (GDI) valuesfor each market. For every i species, we multiplied its rmax

(i.e., rmax i) by the annual number of carcasses (ni). Wethen added up the products computed for all species, andthis sum was divided by the total number of individualcarcasses annually traded in the market (N):

GDI =∑S

i=1 (rmax i ∗ ni)

N, (1)

where rmax i∗ni is derived for species i to S in the mar-ket sample. A higher GDI (range 0–1) indicated a moredepleted supply area.

Species Composition in Markets and Human Density

We examined the relationship between the contributionof the various mammalian groups in the markets and hu-man density within the 15-km buffers. For this, we binned


Fa et al. 5

human density into 3 classes: low (below the first quar-tile), medium (between the first and the third quartiles),and high values (above the third quartile). We appliedanalysis of variance (ANOVA) to the arcsine-transformedpercentage of carcasses (Sokal & Rohlf 1981) of a giventaxonomic order to test for differences. Only the 4 mostrepresentative orders (ungulates, rodents, primates, andcarnivores) were analyzed. Tukey’s honestly significantdifference (HSD) post hoc tests were applied when sig-nificant differences (p < 0.05) were found between thehuman density classes.

We used the same approach to analyze the relation-ship between body size of bushmeat species and humandensity. Species were divided into 3 groups according totheir mean body size (small mammals, <1 kg; medium-sized mammals, 1–10 kg; large mammals, >10 kg). Then,we applied ANOVA and Tukey’s HSD post hoc tests to thearcsine-transformed percentage of carcasses belonging tothe species of a given body-size group.

Spatial Autocorrelation among Market Buffer Areas

We used Moran’s I spatial autocorrelation statistic(Diniz-Filho et al. 2003) to test for autocorrelation in theGDI calculated for all markets, which we derived fromthe overlap between neighboring buffers. To ensure arobust outcome, we tested for autocorrelation 4 timesby increasing distance intervals from markets by 2, 3, 5,and 10 km. All results confirmed a significant positivespatial autocorrelation in GDI of 23–27 km distanceamong markets (p < 0.05, randomization test based on500 permutations).

Correlates of GDI

Because we found evidence of autocorrelation amongoverlapping buffers, we employed the following analyt-ical methods to control for this. We analyzed the rela-tionship between the markets’ GDIs and predictor vari-ables describing environmental conditions (vegetationtypes) and anthropogenic pressures (human population,accessibility, fires, and land use) (Supporting Informa-tion). For this, we used 2 complementary approaches,for which we employed the Spatial Analysis in Macroe-cology (SAM version 4.0) software (Rangel et al. 2010).We first used a conditional autoregressive (CAR) model(Keitt et al. 2002) to analyze the continuous variation ofGDI throughout the study area. Second, we employed theenvironmental favorability function (FF) (Real et al. 2006;Acevedo & Real 2012), based on autologistic regression(Augustin et al. 1996), to analyze the difference betweenmarkets with high GDI values (depleted faunal communi-ties) and markets with low GDI values (better-conservedcatchment faunas). Both CAR models and autologisticregression overcome the impact of autocorrelation onthe relationship between GDI and predictor variables.

CAR operates with distance-based weight matrices thatspecify the strength of interaction between neighboringsites (Keitt et al. 2002). Use of CAR assumes the GDI isalso a function of the GDI values at neighboring locations.We built a multiple CAR model by selecting predictorvariables along a forward Akaike information criterion(AICc)-based stepwise search. Following Burnham andAnderson (2002), we used the AICc for finite samplesizes and stopped when the addition of a new variableinvolved an AICc decrease of <2.

To apply FF, markets were assessed along a gradientof favorability for depletion ranging from 0 to 1, whichdescribes local deviations from the overall probabilityof obtaining depleted fauna in markets. We first esti-mated which GDI value represented a cut-off point alonga transition from nondepletion to depletion (procedureexplained in Supporting Information). The FF was basedon the probability (p) of finding a depleted market fol-lowing the formula in Real et al. (2006) (SupportingInformation). Probability was calculated by performingan autologistic regression model of high and low GDIvalues (1/0) on the predictor variables in Supporting In-formation. Autologistic regression accounts for spatial au-tocorrelation through the addition of an autocovariancevariable, which is defined by a weighted sum of obser-vations in neighboring locations, into the logit equation(Augustin et al. 1996). A multiple autologistic regressionmodel was built employing a forward AICc-based step-wise search. McFadden’s R2 was used to assess modelfit; χ2 to evaluate significance; AICc for parsimony-basedmodel comparison; sensitivity, specificity, and kappa forclassification accuracy; and area under the curve (AUC)for discrimination capacity.

To avoid multicollinearity, the variable list was fil-tered so that coefficients of determination (R2) >0.9 andvariance inflation factors >10 were avoided (Kleinbaumet al. 2007). Correlations between all variables are givenin Supporting Information.

Results

Over 45 mammal taxa (of which 16 primates, 12 un-gulates, 7 carnivores, 4 rodents, 2 pangolins, 1 hyrax,1 elephant, 1 bat, and 1 manatee were identified tospecies) were traded across all markets (Supporting In-formation). The mean number of carcasses traded permarket was 2130.60/year (SE 321.60). Mean biomass was21,398.42 kg/market/year (SE 3400.19). Duikers (Phi-lantomba sp., Cephalophus spp.) and other ungulates(sitatunga [Tragelaphus spekii], bushbuck [Tragelaphusscriptus], buffalo [Syncerus caffer]), and rodents (giantpouched rat [Cricetomys spp.], grasscutter [Thryono-mys spp.], African brush-tailed porcupine [Atherurusafricanus]) comprised most of the mammals sold in eachmarket, 42% and 39%, respectively. Primates (11%) and



Figure 2. Relationship between human population density and bushmeat species traded in markets in theCross-Sanaga Rivers region, Nigeria, and Cameroon for 2002–2003: (a) ungulates, (b) rodents, (c) primates, and(d) carnivores (whiskers, 95% confidence intervals; lines within bars, median). Human density was divided into 3classes: low (below the first quartile), medium (between the first and the third quartiles), and high (above thethird quartile).

carnivores (6%) comprised fewer carcasses whereas pan-golins, hyrax, bats, elephant, and manatee comprised therest (2%).

Human density surrounding the markets ranged fromvery low (3.69 inhabitants/km2) to very high (803.29inhabitants/km2, characteristic of urban and periurbanzones). Mean human density was 97.45/km2 (SE 18.87).Road density was high, averaging around 136.31 m/km2

(SE 5.96) and ranging from 12.85 to 295.24 m/km2.Percentage of agricultural land (cropland and croplandmosaic) was 9.05% (SE 1.48, range 0.00–54.42%). Fre-quency of fires and percentage of cropland mosaic areawere positively correlated. Human density was positivelycorrelated with road densities, percentage of agriculturalland area, and number of fires and was negatively

correlated with percentage of terra firme forest cover. Allof these measures indicate high anthropogenic pressures.

A median of 14 species (range 5–25 species) was tradedannually at each market, corresponding with a medianof 10,268.18 kg/year (range 580–400,000 kg/year) (Sup-porting Information). The median of the mean body massfor all markets was 5.14 kg (range 3.00–14.25 kg) (Sup-porting Information). Game diversity (median 0.85) andgame dominance (median 2.33) of traded species permarket varied little among markets (Supporting Informa-tion). Median GDI value for all markets was 0.41 (range0.303–0.635).

Human density was significantly related to theproportion of ungulates sold in markets (ANOVA: F2,76 =3.38, p < 0.001); their proportion was significantly lower


Fa et al. 7

in highly populated market areas compared with mod-erately populated market areas (HSD = 0.12; p = 0.030;Fig. 2a). The proportion of rodents was related to humandensity (F2,76 = 3.96; p< 0.001), although in contrast toungulates, higher proportions of rodents were associatedwith higher human populations relative to medium levelhuman population densities (HSD = 0.13; p = 0.021;Fig. 2b). The proportion of primates sold in markets alsovaried significantly with increasing human density (F2,76

= 3.55; p = 0.034; Fig. 2c). The proportion of primateswas higher in market areas with low human populationdensities than in market areas with high human popu-lation densities (HSD = 0.07; p = 0.026). For carnivores(F2,76 = 5.90; p = 0.004) the proportion of animalssold was higher in markets surrounded by high humandensities than in those surrounded by medium humandensities (HSD = 0.05; p = 0.007) or low human densities(HSD = 0.06; p = 0.012; Fig. 2d).

The proportion of small-bodied mammals sold inmarkets was positively related to increasing humandensity (F2,76 = 7.96; p = 0.001); their proportionwas significantly higher in areas with higher humanpopulation densities than in those with medium (HSD =0.01; p = 0.004) or low population densities (HSD = 1.67;p = 0.001; Fig. 3a). Lower proportions of large-bodiedmammals occurred in markets surrounded by highhuman population densities than in markets surroundedby medium population densities (F2,76 = 0.03; p = 0.025;HSD = 78.13; p = 0.023; Fig. 3c). Proportions of mid-sizedmammals in markets were weakly related with humandensity (F2,76 = 3.33; p = 0.041; Fig. 3b); there were nosignificant pairwise differences between density classes.

Three variables (total human population, road totallength, and main rivers total length) were eliminatedto avoid high multicolinearity in the analysis of corre-lates of GDI. The CAR model included 3 predictor vari-ables (F2,76 = 10.29; p < 0.001; R2 = 0.53; AICc =−250.60) (Table 1). The anthropogenic factor was rep-resented by road density (ROADDENS) and number ofhuman settlements with >3000 inhabitants (NHS), bothwere positively related with GDI. The environmentalfactor was represented by sparse vegetation and bareareas (SPARVEGBARE) and was also positively relatedwith GDI. This model showed better fit than an ordinaryleast squares (OLS) regression when autocorrelation wasnot considered (F2,76 = 13.98; p < 0.001; R2 = 0.36;AICc = −228.88). The R2 explained by the predictorvariables was 0.29; 0.07 points less than the OLS.

The GDI cut-off distinguishing depleted faunalcommunities from better-conserved catchment faunaswas 0.397 (Supporting Information). This value is almostidentical to the median GDI observed in the studiedmarkets (0.41), which indicates that nearly 50% of themarkets were above and 50% were below the thresholdfor markets to be considered depleted. The most parsimo-nious favorability model describing differences between

Figure 3. Relationship between human density andbody size of bushmeat species traded in markets inthe Cross-Sanaga Rivers region, Nigeria, andCameroon for 2002–2003: (a) small mammals(<1 kg), (b) medium-sized mammals (1–10 kg), and(c) large mammals (>10 kg) (whiskers, 95%confidence intervals; lines within bars, median).Classes of human population densities are defined inFig. 2’s legend.



Table 1. Anthropogenic pressure and environmental variables ac-counting for game depletion around bushmeat markets in the Cross-Sanaga Rivers region, Nigeria, and Cameroon in an autoregressivemodel of the game depletion index (GDI) and in a favorability modeldescribing favorable areas for depletion (i.e., high GDI).

Model and variablea Bb SE Tc p

AutoregressiveROADDENS 0.00041 0.00032 2.208 0.030

NHS 0.020 0.013 1.564 0.122SPARVEGBARE 0.054 0.048 1.132 0.261constant 0.364 0.040 8.986 <0.001

FavorabilityPOPULDENS 0.026 0.013 1.935 0.053

NHS 1.216 0.741 1.641 0.101AC 9.454 2.821 3.352 <0.001constant −6.514 1.706 −3.818 <0.001

aVariables are ranked according to their order of entry in the model.Definitions: ROADDENS, road density; POPULDENS, human popula-tion density; NHS, number of settlements with >3000 inhabitants;SPARVEGBARE, sparse vegetation and bare areas; AC, autocovari-able accounting for interaction between neighboring sites. For de-tailed descriptions of each variable see Supporting Information.bCoefficient.cStudent’s t.

depleted and nondepleted buffers, according to AICc,included 2 variables related to anthropogenic pressures(McFadden’s R2 = 0.41; χ2 = 44.48; p < 0.0001; AICc =71.43; sensitivity: 0.80; specificity: 0.85; kappa = 0.64;AUC = 0.89): human density (POPULDENS) and numberof human settlements with >3000 inhabitants (NHS)(Table 1). This model, which included an autocovariancevariable, showed higher fit, classification accuracy, anddiscrimination capacity and was more parsimoniousthan a favorability model with the same variables exceptautocovariance (McFadden’s R2 = 0.27; χ2 = 29.42; p <

0.0001; AICc = 84.27; sensitivity: 0.65; specificity: 0.82;kappa = 0.45; AUC = 0.81).

When favorability for depletion was plotted againstPOPULDENS (Fig. 4a), 0–0.95 favorability values werefound irrespective of density when there were <110inhabitants/km2. Above this density, all buffers showedfavorability >0.95 and thus faunal depletion. For all un-favorable values of depletion (F < 0.2; Delibes-Mateoset al. 2012), POPULDENS was <70. A plot of favorabil-ity against NHS (Fig. 4b) showed that almost all buffers(except one) with NHS >0 had favorability >0.66 fordepletion. Favorability was always >0.95 when NHS = 2.All unfavorable values for depletion had NHS <0.

Discussion

Across tropical regions worldwide, demographic pres-sures such as high human population densities are strongpredictors of timber and mineral extraction (Karanthet al. 2006), fire frequency (Hudak et al. 2004), and,

more generally, species extinction (Luck et al. 2004).In African moist forests, changing land-use patterns, es-pecially deforestation (Mayaux et al. 2013), and climatechange (Kaeslin et al. 2012) will affect wildlife. However,there is mounting evidence that species are being rapidlyextirpated and protected areas for wildlife affected dueto commercial trade for meat or ivory (Fa & Brown 2009).This is likely to increase in the near future (Abernathy etal. 2013).

Although immediate hunting regulation is vital for thesurvival of the Central African rainforest ecosystems, tar-geting areas likely to be more at risk can be a decisive firststep. We found the first conclusive evidence of linkagesbetween human pressures and bushmeat in markets as ameans of discerning wildlife depletion patterns. This hasnot hitherto been possible in the absence of large-scale,extensive data, and analyses. Our results with an unprece-dented sample of sites covering over 30,000 km2 confirmthat in bushmeat markets in heavily exploited areas(defined by indirect anthropogenic metrics such as highhuman population densities), the percent contribution oflarger-bodied prey (slow-reproducing species), especiallyungulates and primates, to markets is characteristicallylower than the percent contribution of smaller fast-reproducing prey, such as rodents. We also demonstratedthat carnivores (mostly smaller taxa) become morenumerous in markets as areas become more depleted ofbushmeat. This has not been confirmed before.

Research from other sites in Central Africa are con-sistent with our main results, especially the observationthat species hitherto unimportant in the bushmeat tradegain prominence when ungulates become scarce (Wilkie1989); in lightly hunted rural sites, duikers and otherantelopes are the more common prey (Lahm 1993; Noss1998). Although factors such as habitat quality and humanpressures jointly impact wildlife (Fa & Brown 2009), themost parsimonious interpretation of the composition oftraded species in bushmeat markets is the contrastingability of ungulate and rodent populations to recoverfrom hunting. In our study, this was clearly verified bythe difference in rmax between ungulates (0.39 [SE 0.04],n = 12) and the much higher value (0.68 [SE 0.007],n = 3) for rodents. Furthermore, our data indicated that anumber of prominent anthropogenic factors, rather thanenvironmental variables per se, explained the situation ofwildlife in bushmeat market catchments. Our evidence,supported by other studies in African rainforests, showedthat higher road densities are linked to reduced abun-dance of a number of mammal species due to higherhunting pressure. Heavily populated and accessible areashave fewer duikers, forest buffalos, and red river hogs(Potamochoerus porcus) (Laurance et al. 2006).

Our index of faunal depletion is the first mechanismdevised that provides a quantitative benchmark thatcan be used to assess overexploitation of fauna at largespatial scales. Other methods (e.g., the defaunation


Fa et al. 9

Figure 4. Variation in favorability of depletion relative to (a) population density and (b) number of humansettlements with >3000 inhabitants, which explained, according to the favorability model, differences betweendepleted and nondepleted market buffer zones in the Cross-Sanaga Rivers region (dashed lines and numbers inparentheses, critical regions of the plot as discussed in text; insets, maps of the study area; white circles,unfavorable buffers [i.e., a 15-km radius from the market centroid] for faunal depletion, favorability for depletion<0.2, population density always <70 inhabitants/km2 or no settlements with > 3000 inhabitants; pale greycircles, favorability = 0.2–0.95, population density always <110 or 1 highly populated settlement; dark greycircles, in [a] favorability >0.95, includes all buffers with population density >110 inhabitants/km2 and in [b]favorability 0.66–0.95; black circles, favorability >0.95, includes all buffers with 2 highly populated settlements;white dots, markets with unfavorable conditions for depletion, population density <70 inhabitants/km2 or highlypopulated settlements, intermediate favorability caused mostly by the neighborhood of depleted buffers).

index of Giacomini and Galetti [2013]) identify faunasaffected by hunting through matched site comparisons(e.g., in a forest fragment and a nearby protected area asa reference site). We investigated relationships betweenvariables by employing a correlational study across anunprecedentedly large sample of sites. These relation-ships revealed that depleted environments were drivenby anthropogenic pressures, such as the existence ofhuman population densities above 110 inhabitants/km2

or alternatively by the presence of >1 human settlementwith >3000 inhabitants in the surroundings. In contrast,clearly unfavorable areas for depletion were found wherehuman density did not surpass 70 inhabitants/km2 andclose populated settlements were not present. Somemarkets fulfilled these conditions but had medium orhigh favorability for depletion (white points in themaps of Fig. 4). According to our model, this effectcan be attributed to interactions with neighboring sitessubjected to stronger human pressures.

The GDI is akin to the mean trophic level (MTL)indicator of fishery catches (Pauly et al. 1998). Like theMTL, the GDI can be used as a framework for discerningthe status of hunted mammals in a catchment areafrom observations of the composition of species forsale in bushmeat markets. Although our index may bemore complicated than the much simpler MTL (becauseboth environmental and anthropogenic variables

are implicated), the relationships we found between preyprofiles and surrounding anthropogenic pressures werestrong. Using the GDI with the application of favorabilitymodeling, we determined that the situation in the Cross-Sanaga region is of concern because over half of all theareas sampled are probably severely affected by hunting.This is worrying, given that our study carried out in 2003already showed signs of wildlife depletion in a large partof one of this most biodiverse region in Africa. Morerecent information on the status of wildlife in the area isnot available, but there is some evidence that the situationis probably worse. The reapplication of our index to thesame markets in a follow-up study could be useful toassess how much wildlife populations may have changedsince our study. Accompanying research to understandthe causes and consequences of these changes are alsovital. Moreover, wildlife use management in depletedareas could, consequently, have positive effects not onlyon their own faunas, but also on neighboring areas.

Given the trends in human populations and infrastruc-ture growth, our results suggest an impending large-scaledegradation of ecosystem structure in tropical forestregions. However, we suggest that metrics such asthose we derived provide that extra analytical layer tounderstand the impacts of hunting, currently the mostpervasive human activity in large forest blocks, such asin the Congo Basin.



Acknowledgments

We thank W. F. Laurance, J. Brashares, R. Nasi, J. Dempe-wolf, and C. Huang for helpful comments on the paper.This work was supported by funds from the U.K. Gov-ernment’s Darwin Initiative Fund (Project # 162/10/004).P.J.J. acknowledges support from the John EllermanFoundation and the Whitley Trust. A.L.M. was funded bythe FEDER of the European Union and Agencia de ObraPublica-Consejerıa de Obras Publicas Vivienda, Junta deAndalucıa (Project G-GI3000/IDIG).

Supporting Information

Description and data sources of variables used (AppendixS1), game diversity and game dominance calculations(Appendix S2), mammal species traded (Appendix S3),correlates of the GDI (Appendix S4), Spearman corre-lation matrix of GDI and predictor variables (AppendixS5), and characteristics of bushmeat markets (AppendixS6) are available online. The authors are responsible forthe content and functionality of these materials. Queries(other than the absence of the material) should bedirected to the corresponding author.

Literature Cited

Abernethy KA, Coad L, Taylor G, Lee ME, Maisels F. 2013. Extent andecological consequences of hunting in Central African rainforestsin the twenty-first century. Philosophical Transactions of the RoyalSociety 368: 20120303. http://dx.doi.org/10.1098/rstb.2012.0303

Acevedo P, Real, R. 2012. Favourability: concept, distinctive character-istics and potential usefulness. Naturwissenschaften 99:515–522.

Augustin N, Mugglestone M, Buckland S. 1996. An autologistic modelfor the spatial distribution of wildlife. Journal of Applied Ecology33:339–347.

Bicheron P, et al. 2008. GlobCover 2005—products description and val-idation report. Version 2.1. European Space Agency, Paris. Availablefrom http://ionia1.esrin.esa.int/ (accessed August 12, 2014).

Burnham KP, Anderson DR. 2002. Model selection and multimodel in-ference. A practical information-theoretic approach. Springer, NewYork.

Cowlishaw G, Mendelson S, Rowcliffe JM. 2005. Evidence for post-depletion sustainability in a mature bushmeat market. Journal ofApplied Ecology 42:460–468.

Delibes-Mateos M, Farfan MA, Olivero J, Vargas JM. 2012. Impact ofland-use changes on red-legged partridge conservation in the IberianPeninsula. Environmental Conservation 39:337–346.

Diniz-Filho JAF, Bini LM, Hawkins BA. 2003. Spatial autocorrelation andred herrings in geographical ecology. Global Ecology and Biogeog-raphy 12:53–64.

Dirzo R, Miranda A. 1991. Altered patterns of herbivory and diversity inthe forest understory: a case study of the possible consequences ofcontemporary defaunation. Pages 273–287 in Price PW, LewinshonTM, Fernandes GW, Benson WW, editors. Plant–animal interactions:evolutionary ecology in tropical and temperate regions. John Wileyand Sons, New York.

Dupain J, Nackoney J, Vargas JM, Johnson PJ, Farfan MA, Bofaso M, FaJE. 2012. Bushmeat characteristics vary with catchment conditionsin a Congo market. Biological Conservation 146:32–40.

Fa JE. 2000. Hunted animals in Bioko Island, West Africa: sustainabilityand future. Pages 168–198 in Robinson JG, Bennett EL, editors.Hunting for sustainability in tropical forests. Columbia UniversityPress, New York.

Fa JE. 2007. Bushmeat markets—white elephants or red herrings? Pages47–60 in Davies G, Brown D, editors. Bushmeat and livelihoods:wildlife management and poverty reduction. Oxford UniversityPress, Oxford.

Fa JE, Brown D. 2009. Impacts of hunting on mammals in African trop-ical moist forests: a review and synthesis. Mammal Review 39:231–264.

Fa JE, Purvis A. 1997. Body size, diet and population density in Afrotrop-ical forest mammals: a comparison with Neotropical sites. Journalof Animal Ecology 66:98–112.

Fa JE, Seymour S, Dupain J, Amin R, Albrechtsen L, Macdonald D. 2006.Getting to grips with the magnitude of exploitation: Bushmeat inthe Cross-Sanaga Rivers region Nigeria and Cameroon. BiologicalConservation 129:497–510.

Forboseh PF, Sunderland TCH, Eno-Nkua M. 2007. Priority setting forconservation in south-west Cameroon based on large mammal sur-veys. Oryx 41:255–262.

Giacomini HC, Galetti M. 2013. An index for defaunation. BiologicalConservation 163:33–41.

Hudak AT, Fairbanks DHK, Brockett BH. 2004. Trends in firepatterns in a southern African savanna under alternative landuse practices. Agriculture, Ecosystems & Environment 101:307–325.

Kaeslin E, Redmond I, Dudley N. 2012. Wildlife in a changing climate.FAO Forestry Paper 167, FAO, Rome.

Karanth KK, Curran LM, Reuning-Scher JD. 2006. Village size and for-est disturbance in Bhadra wildlife sanctuary, Western Ghats, India.Biological Conservation 128:147–157.

Keitt TH, Bjørnstad ON, Dixon PM, Citron-Pousty S. 2002. Accountingfor spatial pattern when modeling organism environment interac-tions. Ecography 25:616–625.

Kleinbaum DG, Kupper LL, Nizam A, Muller KE. 2007. Applied regres-sion analysis and other multivariable methods. 4th edition. DuxburyPress, Pacific Grove.

Kumpel NF. 2006. Incentives for sustainable hunting of bushmeat in RıoMuni, Equatorial Guinea. PhD dissertation. Imperial College London,University of London and Institute of Zoology, Zoological Society ofLondon.

Lahm S. 1993. Utilization of forest resources and local variation ofwildlife populations in Northeastern Gabon. Pages 383–401 inHladik CM, Hladik A, Linares OF, Pagezy H, Semple A, Hadley M,editors. Tropical forests, people and food: biocultural interactionsand applications to development. UN Educational, Scientific andCultural Organization, Paris.

Laurance WF, Croes BM, Tchignoumba L, Lahm S, Alonso A, Lee ME,Campbell P, Ondzeano C. 2006. Impacts of roads and hunting oncentral-African rainforest mammals. Conservation Biology 20:1251–1261.

Luck GW, Ricketts TH, Daily GC, Imhoff M. 2004. Alleviating spatial con-flict between people and biodiversity. Proceedings of the NationalAcademy of Sciences USA 101:182–186.

Macdonald DW, Johnson PJ, Albrechtsen L, Dutton A, Seymour S, Du-pain J, Hall A, Fa JE. 2011. Association of body mass with price ofbushmeat in Nigeria and Cameroon. Conservation Biology 25:1220–1228.

Macdonald DW, Johnson PJ, Albrechtsen L, Seymour S, Dupain J, HallA, Fa JE. 2012. Bushmeat trade in the Cross–Sanaga rivers region:Evidence for the importance of protected areas. Biological Conser-vation 147:107–114.

Mayaux P, et al. 2013. State and evolution of the African rain-forests between 1990 and 2010. Philosophical Transactionsof the Royal Society, Series B 368, DOI: 10.1098/rstb.2012.0300.


Fa et al. 11

Nasi R, Taber A, Van Vliet N. 2011. Empty forests, empty stomachs?Bushmeat and livelihoods in the Congo and Amazon Basins. Inter-national Forestry Review 13:355–368.

Noss AJ. 1998. The impacts of the cable snare hunting on wildlife pop-ulations in the forest of the Central African Republic. ConservationBiology 12:390–398.

Oates JF, Bergl RA, Linder JM. 2004. Africa’s Gulf of Guinea forests:biodiversity patterns and conservation priorities. Conservation In-ternational, Center for Applied Biodiversity Sciences, Washington,D.C.

Pauly D, Christensen V, Dalsgaard J, Froese R, Torres FJ. 1998. Fishingdown marine food webs. Science 279:860–863.

Rangel TFLVB, Diniz-Filho JAF, Bini LM. 2010. SAM: a comprehensiveapplication for Spatial Analysis in Macroecology. Ecography 33:46–50.

Real R, Barbosa AM, Vargas JM. 2006. Obtaining environmental favoura-bility functions from logistic regression. Environmental EcologicalStatistics 13:237–245.

Rowcliffe JM, Cowlishaw G, Long J. 2003. A model of human hunt-ing impacts in multi-prey communities. Journal of Applied Ecology40:872–889.

Sokal RR, Rohlf FJ. 1981. Biometry. WH Freeman and Company, NewYork.

Vargas JM, Farfan MA, Guerrero JC, Real R. 2004. Caracterizacion de losaprovechamientos cinegeticos a escala macroespacial: un ejemploaplicado a la provincia de Granada (sur de Espana). Ecologıa 18:53–70.

Wilkie DS. 1989. Impact of roadside agriculture on subsistence huntingin the Ituri forest of northeastern Zaire. American Journal of PhysicalAnthropology 78:485–494.


correlates of bushmeat in markets and depletion of wildlife

Documents

wildlife species

bushmeat markets

markets gdi

human density

species composition

mammal species

species highest rmax

human population density